Comparative analysis of gene expression in mammalian claustrum subdivisions

Abstract

The claustrum is a major connectivity hub for cortical circuits and has recently become a nexus for research on cognition, brain state transitions, and mood disorders. While most recent studies focus on mice, there are ongoing debates regarding the definition of rodent claustrum borders, subregions, development, and constitutive cell types. Hence, problems mount when attempting to compare the claustrum across mammals with diverging brain size and morphology. In particular, understanding the primate claustrum remains challenging, because numerous incongruent parcellation schemes for the primate claustrum exist. Here, we propose that evolutionarily conserved subdivisions of the claustrum complex can be identified across mammals by combining gene expression patterns, cytoarchitecture, and topological position. To this end, we compare selected claustrum-specific genes, such as Nurr1, Oprk1, Lxn, and Cdh8, in several species, including rats, Etruscan shrews, tree shrews, marmoset monkeys, and macaque monkeys, to examine patterns of claustrum subdivision across these species.

Significance Statement.

Interest in the structure and function of claustrum is continuously growing, but most recent studies focus exclusively on mice. To better understand its structure across mammals, we used in situ hybridization to study gene expression patterns, using claustrum markers like Nurr1, Oprk1, Lxn, and Cdh8 in the rat. Applying organizational features identified in the rat claustrum complex, we compared gene expression patterns across additional mammalian species. We observed conserved patterns in molecular and spatial organization despite substantial differences in brain size. These findings support the presence of shared structural characteristics and contribute to a comparative understanding of claustrum organization.

Introduction

The mammalian claustrum is an elongated area of gray matter embedded between the insular cortex and the striatum. Historically, its secluded position and intermingled cell types made recording and manipulating claustrum neuron activity particularly difficult, impeding functional studies. This barrier has been circumvented with the advent of transgenic mouse models, which now serve as the primary basis for claustrum research and have revealed a surprising range of claustrum functions in recent years (1–3). Recent studies have emphasized the role of the claustrum in regulating brain states (4–7) but emerging evidence also suggests a role in anxiety and mood disorders (8, 9). However, a significant challenge persists: reconciling findings in mice with the function of the primate and human claustrum (10), because the sophisticated tools enabling progress in mice, such as cell type–specific tracing, remain difficult to replicate in other species (11).

Thus, although the claustrum has been investigated in a wide range of species, a central problem is that it remains unclear whether what we learn about the claustrum in mice can be readily translated to other species. A key prerequisite for translating knowledge of claustrum connectivity and function across species is to understand the basic structure of the claustrum, i.e. its cell type composition, cellular architecture, and gene and protein expression patterns in different species. Like the cerebral cortex, the claustrum consists mainly of glutamatergic neurons. In larger mammals, the claustrum is enclosed by the extreme capsule and the external capsule, two prominent fiber tracts, and thus easily recognized. Rodents and other small mammals lack an easily identifiable extreme capsule, complicating comparisons across species. Furthermore, an ongoing debate aims to comprehensively define the borders of the claustrum complex in rodents (12–17). Rodent brain atlases describe a small dorsal claustrum (dCL), a round, elongated ventral claustrum (vCL) underlying the insular cortex, and a much broader dorsal endopiriform nucleus (DEn) underlying the olfactory cortex (18–20). Most studies in rodents focus on the vCL, yet opinions differ on its relation to the dCL and the DEn, which exhibit differences in connectivity and in function. All three nuclei (dCL, vCL, and DEn) are collectively referred to as the claustrum complex (14), since they share similarities in gene expression (17, 21–24) and neuronal birth dates (25–27), which led to the suggestion that they share a common developmental origin (28). Surprisingly, earlier studies by Arimatsu et al. (29, 30) had shown widespread expression of claustrum-specific markers in deep layers of lateral cortical regions, yet outside the nuclei of claustrum complex, suggesting that there is a population of claustrum-like neurons embedded in the deep layers of the lateral cortex (28). In support of this argument, recent work identified a population of deep layer cortical glutamatergic neurons that are transcriptomically identical and show complementary connectivity to neuronal cell types located in the claustrum (24, 31, 32). In summary, much like the cortex, the claustrum can be divided into subregions that differ in connectivity, cell type composition, and function. However, identifying the same subdivisions of the claustrum complex in other species, particularly in primates, remains challenging (2, 11, 22, 23, 33, 34). Additionally, there is evidence in rodents that claustrum-like cortical (CLC) neurons are present outside the claustrum complex but it remains unknown whether they exist in other species.

Here, we aim to overcome this knowledge gap by comparing the claustrum of several mammals. We focus on gene expression, which is more likely to be conserved across species than other anatomical properties, such as protein expression, myelination patterns, or connectivity (35). Recent studies have identified numerous claustrum-specific genes in mice (17, 22–24, 36–38), most of which are highly distinct from genes expressed in cortical cell types. However, except for a few studies (39–42), little is known about the expression of claustrum-specific gene markers in species other than mice. Furthermore, while many claustrum-specific genes are known, their relation to the subdivisions of the claustrum complex remains tentative. To this end, we used in situ hybridization to examine the expression of several claustrum-specific genes in the rat claustrum to identify the subdivisions and cellular architecture of the claustrum complex. Next, we detailed the regional borders of the claustrum complex and claustrum-like neurons in the rat brain. Finally, we study the same set of claustrum-specific genes in a wider range of species, including Etruscan shrews (Suncus etruscus), tree shrews (Tupaia belangeri chinensis), marmoset monkeys (Callithrix jacchus), and macaque monkeys (Macaca fascicularis). We find that the main subdivisions of the claustrum complex can be identified based on topology and gene expression patterns in all mammals studied, though they vary dramatically in size and shape. Thus, our results suggest shared evolutionary features of claustral organization and offer a basis for comparative studies of claustrum connectivity and function.

Results

Genetic architecture of the rat claustrum complex

To provide a basis for cross-species comparison of gene expression patterns, we first aimed to establish which groups of genes are expressed in similar patterns in the claustrum complex, consisting of the dCL, vCL, and dorsal DEn. Starting from genes listed in the Allen Mouse Brain atlas as claustrum markers (22, 43) we surveyed a range genes using colorimetric in situ hybridization in complete series of the rat forebrain. First, we focused on the forebrain near the anterior commissure and used Nurr1 (Nr4a2) as a general claustrum complex marker gene (Fig. 1A and B). At this anterior–posterior level all conventional claustrum complex subdivisions are readily apparent. We thus label the vCL with a solid line and other structures with dashed lines. We also note that the ventral part of DEn has a lower staining intensity than its dorsal part, which we term vDEn and dDEn (Fig. 1B and C), consistent with earlier proposals that the dorsal DEn can be divided into at least two subregions (25, 26, 29). A schematic of the expression pattern of Nurr1 in the claustrum complex is shown in Fig. 1C. The different expression patterns within the DEn are more clearly seen in second group of genes (Fig. 1D–F). Oprk1 and Lxn show much less expression in vDEn compared to dDEn. In addition, these genes are also densely expressed in the dCL and vCL (Fig. 1B–F). Only a few genes that we surveyed showed expression specific to the dorsal and vCL, but lower or absent expression in the dorsal DEn. In that group, Cdh8 and Sntb1 are interesting because Cdh8 is more strongly expressed in vCL, while Sntb1 labels a subset of cells in dCL and vCL and both genes have low expression levels in DEn (Fig. 1G–I). The DEn is more readily distinguished from dCL and vCL by its more ventral location, size, and numerous differentially expressed genes, including Pcp4 and Nnat (Fig. 1J–L). More specifically, it is the ventral part of the DEn (vDEn) that expresses numerous genes distinct from the dorsal DEn (dDEn) as well as dCL and vCL. Pcp4 is absent from vCL and dCL, but broadly expressed in the striatum and cortical neurons that are wrapped around dCL and vCL (Fig. 1J). Finally, Nnat is strongly expressed in vDEn, and more weakly in dDEn, other parts of the claustrum complex, and the surrounding insular cortex (Fig. 1K). Thus, we classify gene expression patterns similar to Pcp4 and Nnat with higher expression in vDEn than dDEn as the fourth group observed in the claustrum complex (Fig. 1L).

Figure 1.

Classifying expression patterns of claustrum enriched genes in the rat. A, B) Nurr1 expression in the claustrum complex shows major subdivisions: dCL, vCL, and DEn. In DEn, Nurr1 show more expression in dorsal part that we label dDEn, and less expression in ventral part that we label vDen. C) A simplified sketch of Nurr1 expression pattern in claustrum complex showing all subdivision labeled with less staining in vDEn. D, E) Oprk1 and Lxn are frequently cited as claustrum markers. All are expressed in dCL, vCL, and Den; however, in Den, they are much less expressed in vDEn. F) A simplified sketch of Oprk1 and Lxn expression pattern in claustrum complex showing much less staining in vDEn. G, H) Expression of Cdh8 and Sntb1 is largely absent in DEn. Cdh8 is more strongly expressed in vCL, while Sntb1 appears more strongly expressed in dCL. I) A simplified sketch of Cdh8 and Sntb1 expression pattern in claustrum complex showing mainly expression in dCL and vCL. J, K) Pcp4 and Nnat are nearly absent or less expressed in dCL, vCL, and dDEn, but show strong expression in vDEn. L) A simplified sketch of Pcp4 and Nnat expression pattern in claustrum complex showing more expression in vDEn. Scale bar in b applies to rest panels.

Cellular architecture and “core” and “shell” regions of the rat claustrum

Next, we aimed to understand the relation of local cytoarchitecture and gene expression using a subset of genes at the same anterior–posterior coordinates shown in Fig. 1. To this end, we combined neutral red staining (Fig. S1a and b) with in situ hybridization for Lxn, Cdh8, Cpne4, and Pcp4 (Fig. S1c–f). Neutral red staining revealed a number of distinct cytoarchitectonic features of the claustrum complex: The dCL is intermingled with neurons in deep cortical layers and is difficult to delineate by cytoarchitecture alone (Fig. 1A), whereas the vCL region forms a recognizable nucleus below the layers of the insular cortex. We also note a relatively cell sparse region in DEn, marked with an asterisk (Fig. S1a). The vCL contains a “core” of darkly stained cells, surrounded by more lightly stained cells (Fig. S1b). The DEn can be divided into a dorsal (dDEn) and a ventral (vDEn) region. The dDEn is oriented toward the vCL and the olfactory cortex and composed of relatively homogeneously stained cells, whereas the vDEn is located closer to the striatum and consists of more dispersed cells. Ventral to the vDEn, the intermediate endopiriform nucleus (IEn) consists of more densely packed, smaller cells (Fig. S1a), and is considered to have a different developmental origin from the claustrum complex (21). Lxn expression densely labels vCL, dCL, scattered neurons dorsal to dCL, and the dDEn (Fig. S1c). In contrast, Cdh8 labels mainly vCL, with less expression in dCL, and no apparent expression in DEn (Fig. S1d). Cpne4 expression differs from the other genes described here in that it appears to label only a subset of vCL neurons, consistent with the “core” and “shell” organization within the vCL. In addition, Cpne4 is useful for identifying the ventral border of the DEn, as it strongly labels the densely packed IEn neurons (Fig. S1e). Both Cpne4 and Pcp4 are strongly expressed in the vDEn, but Pcp4 is absent from vCL (Fig. S1f).

To further investigate the arrangement of claustrum and DEn cell types, we used double fluorescence in situ hybridization. Combining Nurr1 and Oprk1 expression (Fig. S2a–c), we found that they overlap in the vCL and in the dDEn, while only Nurr1 is expressed in the vDEn. Confirming our previous finding of stronger Cpne4 expression in the central region of vCL, we combined NeuN antibody staining with in situ hybridization for Cpne4. As shown before (Fig. S1), we found that Cpne4 expression is stronger in the central region of vCL, surrounded by a region of weaker Cpne4 expression (Fig. S2d–f). When we examined the co-expression of Nurr1 and Pcp4, we found only little overlap in vCL or dDEn, while there is a large degree of overlap between Nurr1 and Pcp4 in vDEn (Fig. S2g–i).

Anterior–posterior extent of the rat claustrum and distribution of CLC neurons

Next, we asked whether the subdivisions identified in the anterior part of the rat forebrain are constantly present throughout the entire anterior–posterior axis (Fig. 2). As a first step, we focus on Oprk1 (Fig. 2A) and use Nurr1 (Fig. 2B) for comparison. Two main parts of the claustrum complex (dCL and vCL) are easily recognized in the anterior half of the rat forebrain in Oprk1-stained sections (Fig. 2A). The difference in the expression pattern of Nurr1 and Oprk1 in the DEn are consistently present from anterior to posterior (Fig. 2A and B). The anterior border of vCL is identified by a combination of Nurr1 and Cdh8 (Fig. 2C and D). The differential gene expression in the DEn is also shown by Nurr1 and Pcp4 (Fig. 2E and F). In addition, Oprk1, Nurr1, and Lxn are expressed in a dispersed population of cells in the deeper layers of the lateral and posterior neocortex, which we label as CLC neurons (Fig. 2A, B, G, and H) (32).

Figure 2.

Anterior–posterior extent of claustrum subregions in the rat. A) Anterior–posterior series of coronal sections stained for Oprk1. B) Anterior–posterior series of coronal sections stained for Nurr1. The cell populations of the dCL appear more like a defined nucleus in anterior sections, whereas in more posterior sections, a seemingly continuous but much more scattered population of cells appears. These deep layer claustrum-like neurons are similar to other claustrum neurons in terms of development, gene expression, and connectivity. The vCL in the rat is only present in the anterior half of the cortical hemisphere, while the DEn, particularly its ventral subregion continues throughout the entire anterior–posterior axis. Inset in a indicates the plane of coronal sections on an outline of the brain. dCL, dorsal claustrum; vCL, ventral claustrum; dDEn, dorsal part of dorsal endopiriform nucleus; vDEn, ventral part of dorsal endopiriform nucleus. Boxes indicate positions of higher resolution images in c–h. C, D) Nurr1 and Cdh8 staining identify vCL in the anterior part of claustrum. E, F) Nurr1 and Pcp4 have different expression in the ventral part of DEn, showing dorsal and ventral part of DEn. G, H) Nurr1 and Lxn label the CLC deep layers neurons. Scale bar in b applies also applies to a. Scale bar in h applies also applies to c–f.

We also collected a more complete set of serial sections to show the claustrum subdivisions constantly present in anterior to posterior sections, including Nurr1, Lxn, Cdh8, Pcp4, and Cpne4 (Fig. S3). The anterior border of the vCL can be identified by the absence of obvious Cdh8 expression in more anterior sections. (Fig. S3c1 and c2). The vCL, as shown by the expression of Oprk1, Nurr1, Lxn, Cdh8, and Cpne4 becomes gradually smaller toward the posterior part of the brain and disappears near the beginning of the dorsal hippocampus (Fig. S3a8–e8). In contrast, while the dCL appears as a distinct nucleus in anterior sections, it becomes increasingly dispersed in more posterior sections. The Nurr1- and Lxn-labeled CLC neurons appear as an even more widely dispersed population of neurons (Fig. S3a9–11 and b9–11). Comparing Nurr1 and Lxn expression in the DEn along the anterior–posterior axis, we observe that the Oprk1-positive dDEn gradually becomes smaller in the posterior cortex, whereas the Nurr1-positive vDEn is present along the entire anterior–posterior axis. In the most posterior sections shown here (Fig. S3a9–11 and b9–11), the vDEn appears to merge with the deep layers of the lateral entorhinal cortex. Generally, Nurr1 is expressed more broadly and also densely labels the dCL, vCL, and DEn, where we can use it in comparison with Oprk1 to distinguish dDEn and vDEn.

Comparative anatomy of the Etruscan shrew claustrum

After describing the architecture of the claustrum complex and the transcriptomically related CLC neurons in the rat, we explored the cellular architecture of the claustrum in other species beginning with the smallest mammal, the Etruscan shrew (Fig. 3). Using in situ hybridization for Oprk1 and Nurr1, we identify subregions similar to those in the rat, including a central, compact region (vCL) and a more dorsal region (dCL) whose cells appear to be increasingly scattered toward the posterior cortex. More posteriorly, we observe a dispersed group of cells in a similar topographic position as CLC neurons in the rat (Fig. 3A and B). Unlike in the rat, Oprk1 is also prominently expressed in L5 neurons of the cortex. However, their different staining intensity and comparison with Nurr1 reveals that these L5 neurons are distinct from CLC neurons (Fig. 3A and B). As in the rat, the DEn is located deep to the olfactory cortex and expresses Oprk1 in a more dorsal subregion, whereas Nurr1 is expressed more broadly (left box in Fig. 3B). This suggests that a similar subdivision of the DEn might be present in the Etruscan shrew as in the rat, so we prepared sections with antibody staining for Pcp4 (which labels vDEn in the rat; Fig. 3C and D). However, while Pcp4 was somewhat enriched in the DEn, there was no clear subdivision into two parts. At the same time, Pcp4 is expressed in different layers of the surrounding cortical regions that wrap around the dCL and vCL nuclei more clearly in the Etruscan shrew than in the rat (Fig. 3C and D). To specifically label vCL, we compared Nurr1 and Cdh8 expression (Fig. 3E and F). Previous authors identified a special layer underneath the vCL in shrews and related small mammals termed “lamina profunda (LP),” suggesting that it is a part of the claustrum (44). However, we find that markers of the claustrum core region are absent in the LP (Fig. 3E and F); see also Pcp4 expression in the LP (Fig. 3C and D). Interestingly, ventral to the CLC neurons is a further group of cells positive for both Oprk1 and Nurr1, that we term white matter claustrum-like neurons (WM^CL, Fig. 3A, right box in Fig. 3B). To investigate these populations in more detail, we show higher magnification views stained for Nurr1 and Synpr (Fig. 3G and H). The expression of Synpr in the rat claustrum complex is highly similar to that of Oprk1 and Lxn (Fig. 1D and E; data not shown). Both CLC neurons and WM^CL neurons express Nurr1, Synpr, Oprk1, and other claustrum marker genes (e.g. Ntng2, data not shown; Fig. 3G and H). By position and gene expression pattern, Etruscan shrew CLC neurons are most likely similar to the CLC neurons in the rodent posterior and lateral cortex. However, in the Etruscan shrew, CLC neurons are predominantly located in layer 5 and to a lesser extent in layer 4 and layer 6 (Fig. 3G), rather than mainly in layer 6 as in rodents, indicating some variability in their laminar position across species. A second, unusual population of cells positive for all tested claustrum-specific marker genes is embedded within the white matter (WM^CL neurons) Counterstaining with neutral red reveals that WM^CL neurons are indeed embedded within the white matter between cortex and striatum. WM^CL differ from subplate neurons in that they express other claustrum markers such as Oprk1 and Synpr (Fig. 3A), which we did not observe in rat subplate cells. Another difference is that subplate neurons are located in the deep part of cortical layer 6, whereas WM^CL neurons are embedded in the white matter. We have not observed a similar population of white matter embedded claustrum-like neurons in rodents.

Figure 3.

Claustrum-specific gene expression in the Etruscan shrew. A) Anterior–posterior series of coronal sections stained for Oprk1. B) Anterior–posterior series of coronal sections stained for Nurr1. Boxes indicate positions of higher resolution images in c–h. C, D) Pcp4 antibody staining (green) at the position indicated in b leaves the vCL and dCL regions unstained, while showing expression in DEn and in cortical neurons above and below (LP) the claustrum complex. DAPI counterstaining shown in (purple). E, F) Nurr1 and Cdh8 staining at the same location identifies subdivisions of the claustrum complex. Cdh8 is absent from DEn and weakly expressed in dCL as in rats. G, H) Nurr1 and Synpr staining at a more posterior location in the cortex reveals two populations of claustrum-like cells embedded in the CLC deep layers of cortex and the white matter (WM^CL). dCL, dorsal claustrum; vCL, ventral claustrum; DEn, dorsal endopiriform nucleus. Scale bar in b applies also applies to a. Scale bar in f applies to c–h.

Comparative anatomy of the tree shrew claustrum

Although Etruscan shrews and tree shrews share part of their name, they are not closely related. Instead, tree shrews are closely related to both rodents and primates, although the exact phylogenetic position is debated (45). To explore gene expression patterns in the tree shrew claustrum, we stained brain sections with probes against Oprk1, Lxn, Nurr1, Cdh8, and Synpr (Fig. 4). Comparing Oprk1 and Nurr1 allowed us to identify the major subdivisions of the claustrum complex (Fig. 4A and B). While dCL is typically smaller than vCL in coronal sections of rat brains, this is not necessarily true in other species. In the Etruscan shrew, we already found that dCL is larger than vCL in some sections (Fig. 3). In the tree shrew, we observe a similar phenomenon, as we identify the DEn as the Oprk1- and Nurr1-positive nucleus embedded within the olfactory cortex, the vCL as the adjacent Oprk1- and Nurr1-positive nucleus, and the dCL as the most dorsal Oprk1- and Nurr1-positive nucleus (Fig. 4A, left box in Fig. 4B). This is confirmed by comparing Cdh8 and Nurr1 staining in higher magnification views (Fig. 4C and D), which show dense staining of Cdh8 in the vCL, as in the rat (Fig. 1G). Similarly, Nurr1 and Lxn expression show striking differences in DEn, with Nurr1 being more widely expressed. Based on these patterns, we designate the more lateral and dorsal part as dDEn and the more medial and ventral part as vDEn (Fig. 4E and F). Interestingly, we did not find a clear expression of any of the tested claustrum marker genes embedded within the cerebral cortex comparable to the CLC neurons in the rat. Instead we observed that the dCL is continuously present throughout the anterior–posterior axis and actually appears to be larger in the posterior parts of the tree shrew brain (Fig. 4A and B). A higher magnification view of Synpr and Cdh8 staining (Fig. 4G and H) shows that the dorsal part is patterned by diagonal lines, as recently observed in the mouse dCL (14). Similar to more anterior sections, Cdh8 is more strongly expressed in a small longitudinal subregion that we designate as vCL, while Synpr, Nurr1, and Lxn are equally strongly expressed in both parts (Fig. 4G and H). In summary, the use of claustrum-specific gene markers in tree shrews reveals a number of similarities and differences. In tree shrews: (i) the dCL is a large, continuous structure; (ii) the dorsal and vCL can be distinguished by Cdh8 expression as in Etruscan shrews and rats; and (iii) in contrast, CLC neurons were not obvious with the tested gene markers.

Figure 4.

Claustrum-specific gene expression in the Chinese tree shrew. A) Anterior–posterior series of coronal sections stained for Oprk1. B) Anterior–posterior series of coronal sections stained for Nurr1. Boxes indicate positions of higher resolution images in c–h. C, D) Nurr1 and Cdh8 staining in the anterior part of the tree shrew brain identify dCL, vCL claustrum complex. Cdh8 is absent from DEn and weakly expressed in dCL as in rats. E, F) Nurr1 and Lxn expression in a more posterior section show differences in DEn, indicating a similar division into dDEn and vDEn as in rats. G, H) Synpr and Cdh8 staining in the posterior part of the claustrum complex show continuous differences in Cdh8 expression and other claustrum markers, indicating a continuous subdivision into dCL and vCL. dCL, dorsal claustrum; vCL, ventral claustrum; dDEn, dorsal part of dorsal endopiriform nucleus; vDEn, ventral part of dorsal endopiriform nucleus. Scale bar in b applies also applies to a. Scale bar in d applies also applies to c. Scale bar in f applies also applies to e. Scale bar in h applies also applies to g.

Comparative anatomy of the marmoset claustrum

Turning to primates, we first investigate gene expression using claustrum-specific markers in neonatal marmoset monkeys using images from the gene expression database introduced by Shimogori et al. (46) and Kita et al. (47). Oprk1 and Nurr1 expression is already well-developed at birth in the marmoset (Fig. 5A and B). In contrast to most primate brain atlases, we designate dCL and vCL as separate subregions of the claustrum complex. As in tree shrews, the dCL is a large subregion of the claustrum complex, while we identify the DEn as a relatively small region close to the olfactory cortex, consistent with the relatively smaller size of the olfactory cortical regions in primates. This interpretation is supported by the gene expression patterns of Cdh8, which—as in the other species investigated above—is strongly expressed in vCL and more weakly in dCL, and Sntb1, which shows the opposite pattern (Fig. 5C and D). In contrast, Cdh8 and Sntb1 are absent from the DEn, which instead strongly expresses Nurr1 and Nnat. In addition, a subset of cells in the DEn weakly express Oprk1 and Nnat, suggesting a subdivision into dDEn and vDEn as in rats and tree shrews (Fig. 5E and F). In the posterior regions of the primate claustrum, subdivisions are much more difficult to discern, due to the bending of the anterior–posterior axis in primates. Interestingly, however, Cdh8 and Sntb1 continue to be expressed in different compartments, tentatively identifying a vCL subdivision in the posterior claustrum (Fig. 5G and H).

Figure 5.

Claustrum-specific gene expression in the Marmoset monkey. A) Anterior–posterior series of coronal sections stained for Oprk1. B) Anterior–posterior series of coronal sections stained for Nurr1. Boxes indicate positions of higher resolution images in c–h. C, D) Cdh8 and Sntb1 staining near the anterior commissure identify dCL and vCL of the claustrum complex. Cdh8 is more strongly expressed in vCL and more weakly in dCL while Sntb1 is more strongly expressed in dCL and more weakly in vCL. Both are absent from DEn as in the rat. E, F) Near the anterior commissure Oprk1 is expressed in dCL, vCL, and part of DEn, while Nnat has strong expression in a subset of DEn neurons and weaker expression in other parts of the claustrum complex, suggesting a subdivision of DEn into two parts as in the rat. G, H) Cdh8 and Sntb1 expression in the posterior part of the claustrum show a similar alternating pattern of staining albeit in a different spatial arrangement, similar to the pattern observed in tree shrews. dCL, dorsal claustrum; vCL, ventral claustrum; dDEn, dorsal part of dorsal endopiriform nucleus; vDEn, ventral part of dorsal endopiriform nucleus. Scale bar in b applies also applies to a. Scale bar in h applies to c–h.

Comparative anatomy of the macaque claustrum

Macaques are more closely related to humans than marmosets. Their cerebral cortex shows more extensive folding and the shape of the claustrum complex is less recognizably similar to the species we previously studied. We obtained tissue samples from a juvenile macaque monkey and stained them for Oprk1 and Nurr1. As in the rat, Oprk1 and Nurr1 are expressed along the entire anterior–posterior extent in the macaque claustrum complex (Fig. 6A and B). Staining in the most anterior sections frequently was interrupted by a striated pattern characteristic of the dCL as described above (Fig. 6C and D). In some anterior sections, Nurr1 showed a broader expression pattern than Oprk1 on the ventral side of the claustrum complex, suggesting a subdivision of the DEn as described above in other species (Fig. 6E and F). At about 1/3 of the anterior–posterior extent, just after the appearance of the temporal lobe in coronal sections, the claustrum attains a highly complex shape, with a thin dorsal part extending along the insular lobe and a very broad pool of claustrum cells near the level of the anterior commissure (Fig. 6G and H). Consistent with our data in tree shrews and marmoset monkeys and previous studies (42), we found little evidence in the macaque for a population of cells similar to the rodent CLC neurons extending into the middle layers of the temporal and posterior neocortex.

Figure 6.

Claustrum-specific gene expression in the Macaque monkey. A) Anterior–posterior series of coronal sections stained for Oprk1. B) Anterior–posterior series of coronal sections stained for Nurr1. Boxes indicate positions of higher resolution images in c–h. C, D) Higher resolution views of the anterior claustrum show striated pattern similar to that seen in the tree shrew. E, F) Higher resolution views near the anterior commissure show differential staining of Oprk1 (e) and Nurr1 (f), suggesting similar subdivisions of the DEn as in other mammals studied here. G, H) Higher resolution views in more posterior parts of the claustrum complex illustrate the increasing complexity of claustrum structure. dCL, dorsal claustrum; vCL, ventral claustrum; dDEn, dorsal part of dorsal endopiriform nucleus; vDEn, ventral part of dorsal endopiriform nucleus. Scale bar in b applies also applies to a. Scale bar in g applies to c–h.

Finally, we consider the topological organization of the claustrum complex and provide an overview of Oprk1 expression in five mammalian species (Fig. 7A). Oprk1 expression in mammals has been described above, providing an extensive dataset for cross-species comparisons. Despite the limitations of using coronal sections for comparative gene expression studies, we find a conserved topological relationship of the piriform cortex, claustrum complex, and striatum across five mammalian species (Fig. 7B). While the relative size of the claustrum components and the topographical position of the piriform cortex and striatum vary considerably with brain size and the expanding neocortex in mammals and primates, the neighborhood relations of the claustrum subdivisions remain conserved.

Figure 7.

Topology of Oprk1 expression in the claustrum complex in six species. a) Oprk1 expression in the Etruscan shrew, rat, Chinese tree shrew, marmoset monkey, and macaque monkey show conserved topology and subdivisions of the claustrum complex. Divergence times based on Fan et al. (45). b) The olfactory cortex and striatum show varying topographic yet consistent topological position in relation to the claustrum complex. dCL, dorsal claustrum; vCL, ventral claustrum; dDEn, dorsal part of dorsal endopiriform nucleus; vDEn, ventral part of dorsal endopiriform nucleus; Pir, piriform (olfactory) cortex; Str, striatum; MYA, million years ago.

Discussion

We have described the comparative architecture of the claustrum complex in several mammalian species using a small set of claustrum-specific gene markers. Our findings indicate that the three major subdivisions of the claustrum complex defined in rodents (dCL, vCL, and DEn) are conserved across the species examined.

Cytoarchitecture, gene expression, and nomenclature of the claustrum complex

Many authors have examined the cellular architecture of the claustrum complex in mammals besides rodents (44, 48–51). These studies have found marked variability in the topographic position and number of claustrum complex subdivisions. However, identifying the same subdivisions across species remains challenging when using only cyto- or myeloarchitectonic features. Additionally, some markers that do allow to identify subregions in rodents, such as parvalbumin and calretinin (52, 53) do not necessarily exhibit subregional specificity in other species (54, 55). Therefore, we focus on principal neuron marker genes that broadly label the claustrum complex across species, including Lxn, Oprk1, and Nurr1 as well as markers for specific claustrum cell types or subregions, such as Cdh8, Sntb1, and Pcp4 (29, 38, 43, 53, 56–58). While the gene expression data for the mouse claustrum is extensive, the available data in primates and humans are still limited (11, 40, 42, 59–61) and requires further study, in particular regarding genes that can identify different subregions of the claustrum complex across species.

Another complication for comparative studies of the claustrum complex arises from the fact that terminology that is used differently across species. In bats, the claustrum complex comprises the claustrum (Cl) and the DEn. Within the Cl, core–shell patterns have been described based on the immunoreactivity of Latexin, calbindin, calretinin, and parvalbumin. However, a further subdivision of the Cl into dorsal and vCL has not been reported (41, 58, 62). In primates, the “dCL” is considered equivalent to the combined dorsal and vCL in rodents (54), so some primate brain atlases therefore do not distinguish between the dorsal and vCL (46, 63). A more recent atlas of the primate claustrum (11) introduces four different subdivisions of the claustrum (dorsal, ventral, anterior, and posterior), albeit without clearly defined borders, so that the dCL in (11) is unlikely to be comparable to the dCL in previous studies of either primates or rodents. In humans, the claustrum is typically subdivided into a dorsal (insular) component and a ventral (temporal) component. The dorsal component appears more compact, while the ventral component is more fragmented (64–68). In other large-brained mammals, such as dolphins, whales, and elephants, the claustrum is organized into fragmented claustral islands (69–71). This makes defining the subdivisions of the claustrum complex more challenging. Further studies are needed to identify claustrum subdivisions in large-brained mammals that are comparable to those in mammals with smaller brains. In conclusion, while there is an emerging consensus on the definition of the rodent claustrum complex (14, 18–20, 22, 23), a comparable scheme for other mammals was previously lacking. Our results support the presence of conserved aspects of claustrum organization revealed by comparative gene expression analyses. However, a comprehensive comparative framework will require further investigation, particularly in primates and large-brained mammals.

The vCL and its core and shell regions

Madden et al. (2) recently presented a model for comparing the rodent, primate, and human claustrum complex. This model suggests that the vCL of rodents is equivalent to the entire claustrum region of primates (2). This is consistent with the view that, in mice, the claustrum constitutes a single, densely packed nucleus within the agranular insular cortex (20, 43). We take a different view. Since neurons in the rodent dCL, vCL, and dDEn share highly similar gene expression profiles, we argue that the entire claustrum complex should be considered when defining the claustrum in other mammals. A further aspect that has been difficult to compare across species is the concept of “core” and “shell” regions of the claustrum, which were initially defined using antibody staining for calretinin and parvalbumin (52, 53). In brief, some studies propose that the claustrum “shell” contains cortical cell types surrounding the “core,” which contains claustrum cell types (11, 36). However, a more plausible definition, based on connectivity and gene expression patterns (17, 24, 72), divides the vCL into “core” and “shell” regions identified by two distinct claustrum cell types, while excluding any cortical cell types. For instance, we recently demonstrated that retrograde tracing from the prelimbic, entorhinal, and retrosplenial cortex labels nonoverlapping subsets of Oprk1-positive vCL neurons (32). In macaques, claustrum cell types projecting to the entorhinal cortex are also located in a ventral region of the main nucleus of the claustrum (11), suggesting a conserved topography of projection patterns. However, comparing claustrum connectivity across species remains challenging. For example, studies in mice and cats have suggested that claustrum neurons integrate over a wide range of sensory inputs (73–75), whereas other studies in mice and primates have argued that claustrum neurons preferentially respond to single sensory modalities, which is more consistent with nonoverlapping connections to sensory cortical areas (76–79).

The dCL and CLC neurons

Unlike most previous studies, we examined the dCL and CLC neurons across species. In rodents, the dCL and CLC neurons consist of the same cell types as the vCL and the dDEn (24, 31, 32). Based on their location in the cortex, we propose that CLC neurons could be considered as a posterior extension of the dCL, consistent with their similarity in birth dates that we described previously (25). We recently investigated the connectivity and activity of CLC neurons in comparison with claustrum neurons (32). Most interestingly, we found that CLC neurons exhibit a complementary cortical connectivity profile when compared with vCL neurons. CLC neurons preferentially project to sensory cortices, whereas vCL neurons preferentially project to prefrontal and midline cortices (32). We did not detect an obvious equivalent of CLC neurons in tree shrews or primates. Therefore, we speculate that the primate equivalent of the rodent CLC neurons is likely to be part of the primate dCL region. This is consistent with the connections of the primate dCL to sensory cortical regions but further study is required. In the Etruscan shrew, the mammal with the smallest cortex (80, 81), the claustrum complex is organized similarly to that of rodents. For example, posterior to the dCL, we find a population of cells that are similar to rodent CLC neurons. This finding is consistent with previous studies in small mammals that used the claustrum-specific marker latexin (39, 41). In these studies, the authors also illustrated a large population of latexin-positive cells embedded in the deep layers of the lateral neocortex. In summary, we propose that the dCL is likely a conserved component of the claustrum complex in all mammals, albeit with substantial variation in size and cell type composition. We speculate that, during development, CLC neurons became separated from the dCL in small mammals, but not in larger mammals, such as tree shrews and primates.

The dorsal endopiriform nucleus

We find that the DEn can be divided into two parts: one that expresses claustrum marker genes (dDEn) and another that likely has a different developmental origin (vDEn). Previous studies by our group and others have proposed a similar subdivision of the DEn based on differences in birth dates and anatomical markers (14, 21, 25, 26, 29, 82). However, Grimstvedt et al. positioned the border between the DEn subdivisions differently, potentially due to species variations or the low cell density and high myelination of a portion of the DEn. Previous studies and atlases show a large amount of variation in the delineation of the DEn and its subregions, particularly in primates (11, 22, 33, 46, 83). For example, Pham et al. (83) describe a round region within the marmoset claustrum with low myelination. However, based on gene expression data, and in line with the low density of myelin basic protein in the mouse vCL (14), we argue that this region is the marmoset monkey equivalent of the vCL.

The claustrum in large mammals and claustrum scaling rules

In addition to the difficulties of identifying the same subregions of the claustrum complex in different species, the expansion of the mammalian brain during evolution introduces further complications for comparative studies of the claustrum complex. In a prescient review, Binks et al. (33) discuss the effects of the expanding cerebral cortex on the shape of the claustrum complex, suggesting that, in primates, the dCL would expand and wrap around the vCL, consistent with the data presented here in tree shrews and monkeys. In addition, several authors have studied the claustrum complex in large mammals and identified isolated clusters of claustrum neurons, mostly located anteriorly and dorsally (51, 69–71). Interestingly, a few studies also described clusters of claustrum neurons in the human brain (64, 68). Here, we propose that this pattern may reflect the striated myelination pattern observed in the dCL in mice (14), in dorsal parts of the claustrum complex in marmosets (83), and in the dCL of tree shrews as described in this study.

A previous study found that total claustrum volume increases from 0.12 mm³ in one species of shrews to 580 mm³ in humans (84). Thus, in small mammals, the claustrum makes up ∼1% of the total volume of a cortical hemisphere, while in humans, the claustrum constitutes ∼0.25% of hemisphere volume. Kowiański et al. (84) also find that the volume of the dorsal part of the claustrum complex increases relative to the volume of the ventral parts of the claustrum complex with increasing brain size. In contrast, a recent estimate of the relative volume of claustrum complex subregions in the mouse puts the size of the DEn at three times the volume of the vCL (20). In our recent study of the elephant claustrum, we found that, despite its fragmented, island-like organization, the volume of the elephant claustrum conforms to known mammalian cortico-claustral scaling laws (71). While more studies of claustrum volume are needed that take into account recent definitions of claustrum complex subregions, we can infer from the available data that the dorsal and vCL scale more closely with neocortical regions (to which they are also more strongly connected), while the DEn scales more closely with the olfactory cortex in which it is embedded. This is consistent with the observation that the DEn is relatively larger in small mammals, whereas the dorsal and vCL are relatively larger in large mammals.

Conserved Oprk1 expression in the mammalian claustrum and potential function

Oprk1 encodes the kappa opioid receptor, which has numerous important roles in the nervous system including the modulation of mood, reward, and pain (85, 86). Activating kappa opioid receptors initiates a second messenger cascade that leads to a decrease in cellular excitability (87, 88). Given the widespread expression of Oprk1 in claustrum neurons, the inhibitory effects of activating kappa opioid receptors, and the extensive cortical connectivity of claustrum neurons, specific kappa opioid receptor agonists could have broad modulatory effects on cortical states mediated by claustrum neurons (37, 89–91). The expression patterns of Oprk1 have been mainly studied in rodents, with some limited data available for the human claustrum (92–94), while primate data mainly come from ligand-receptor binding assays (95–98). Interestingly, we find that Oprk1 expression in the claustrum is remarkably conserved across mammals. In contrast, Oprk1 expression in cortex and striatum neurons appears to be more variable across species. For example, there is relatively low expression of Oprk1 in the cortex and striatum of rats, while the Etruscan shrew shows more distinct Oprk1 expression patterns in these regions. Because the kappa opioid receptor and its endogenous ligand dynorphin are implicated in a wide variety of neuropsychiatric disorders, including depression and anxiety (99–101), it is tempting to speculate about the role of the claustrum in these disorders. Recent experimental studies in mice have increasingly highlighted the role of the claustrum in anxiety, depressive-like behaviors, and hallucinogenic states (8, 9, 102, 103). In addition, meta-analyses of fMRI data in human patients also point to a role for the claustrum in mood disorders (104, 105).

In summary, we present comprehensive in situ hybridization data for Oprk1, Nurr1, and other genes in the claustrum of various mammalian species, outlining conserved subregional patterns. While the function of the claustrum remains a complex puzzle, its role in emotional behaviors and disorders merits further study.

Materials and methods

Animals

All experimental procedures were performed according to the institutional guidelines on animal welfare and approved by the local institution in charge of experiments using animals as detailed below. Female and male adult (n = 8, 8–12 weeks old) SD rats were obtained from Beijing Vital River Laboratory Animal Technology Co., Ltd and were used under the following permit: Animal Care and Use Committee at the Shenzhen Institute of Advanced Technology (SIAT), Chinese Academy of Sciences (CAS), China; permit number SIAT-IACUC-20210607-NS-NBJZX-ROBERT NAUMANN-A1706-01. Macaque monkey brain tissue was collected from a 10-month-old male crab-eating macaque (n = 1, body weight 1 kg; obtained from Guangdong Landau Biotechnology Co., Ltd), that was sacrificed for an unrelated experiment (Animal Care and Use Committee at the SIAT, CAS, China; permit number SIAT-IACUC-210326-NS-WH-A1881). Etruscan shrews (S. etruscus) were bred and housed, as previously described (106). Etruscan shrew brain tissue from a previous study (107) was used in this study (Landesamt für Gesundheit und Soziales Berlin, Germany; permit numbers: G0170/15, T0160/14 and T0078/16). Chinese tree shrew (T. belangeri chinensis) brain tissue (n = 4, 2–3 years old) was obtained from the Kunming Institute of Zoology (KIZ), CAS under the permit KIZ-IACUC-TE-2021-06-001. Marmoset monkey (C. jacchus) in situ hybridization images were downloaded from the Marmoset Gene Atlas (https://gene-atlas.brainminds.jp/), further details are described in (46, 47).

Tissue preparation

Animals were anaesthetized by isoflurane, and then euthanized by an intraperitoneal injection of an overdose of pentobarbital. Animals were then perfused transcardially with first 0.1 M phosphate buffered saline (PBS) solution, followed by 4% formaldehyde, from paraformaldehyde in 0.1 M phosphate buffer (PFA). Animals brains were dissected, postfixed and cryoprotected in PBS with 30% sucrose. Free-floating sections were cut with a freezing-microtome.

In situ hybridization

In situ hybridization was performed as described before (107) 108). DIG-labeled riboprobes were used for hybridization on 40–60 µm free-floating cryosections. Hybridization was performed overnight at 60–65 °C. Sections were washed at 65 °C twice in 2 × SSC/50% Formamide/0.1% N-lauroylsarcosine, twice in 2 × SSC/0.1% N-lauroylsarcosine at 37 °C for 20 min and twice in 0.2 × SSC/0.1% N-lauroylsarcosine at 37 °C for 20 min. Sections were blocked in MABT/10% goat serum/1% Blocking reagent (Roche, cat# 11096176001), incubated overnight with sheep anti-DIG-AP (1:1,000, Roche cat# 11093274910). After washing, staining was performed using NBT/BCIP (Sigma Aldrich) in NTMT until satisfactory intensity was reached. Staining reaction was stopped with 10 mM EDTA. Sections were washed, dehydrated and mounted with Eukitt Quick-hardening mounting medium (Sigma Aldrich).

Double fluorescent in situ hybridization was performed using sequential TSA amplification to detect DIG- and FITC-labeled probes. After hybridization and washing, sections were first incubated with sheep anti-DIG-POD (1:1,000, Roche Cat# 11207733910) and TSA was performed using biotin-tyramide (ApexBIO, Shanghai, China). Subsequently sections were incubated with Streptavidin-Cy2 (Jackson ImmunoResearch) to detect the DIG-labeled probe. After inactivating the anti-DIG-POD, sections were incubated with anti-FITC-POD (1:2,000, Roche Cat #11426346910) and TSA was performed using Cy3-tyramide (ApexBIO, Shanghai, China).

Genomic DNA (gDNA) was extracted from liver. cDNA was synthesized from total brain RNA using EasyScript First-Strand cDNA Synthesis SuperMix (Transgen Biotech). Desired DNA fragments were amplified by PCR with the primer pairs indicated in Table S1 (Phusion, NEB). Human gDNA was extracted from HEK293 cells. PCR fragments were individually cloned in pEASY-Blunt Zero backbone (Transgen Biotech) and verified by sequencing. Antisense digoxigenin- and FITC-labeled riboprobes were synthesized according to the protocol recommended by the manufacturer (Roche Cat# 11277073910) and purified with MagicPure RNA Beads (Transgen Biotech). Some sections were counterstained with neutral red. Etruscan shrew and Chinese tree shrew Orpk1 DNA sequences were synthesized by Genewiz.

Some genes such as Nurr1, Cdh8, or Synpr are broadly conserved across species; therefore, we were able to use the probe generated for rats also in other species such as the Etruscan shrew or the Chinese tree shrew. We used a human OPRK1 probe to stain monkey tissue. To avoid false positive results due to using rat ISH probes in different species, we confirmed that the probes used for other species had at least 90% sequence similarity using NCBI or the preliminary sequencing results of the Etruscan shrew genome (107). In all experiments, ISH produced consistent results across multiple sections or multiple animals for each species.

Antibodies

Sections were processed for histochemistry and immunohistochemistry as described previously (109). First, we incubated sections in blocking buffer consisting of 0.1 M PBS and 5% bovine serum albumin (BSA) for 1 h at room temperature. Subsequently, free-floating sections were incubated overnight with the primary antibodies in 0.1 M PBS containing 1% BSA and 0.5% Triton-X (PBS-X). We used primary antibodies against NeuN (Servicebio; 1:1,000, GB11138) and Purkinje cell protein 4 (Sigma; 1:500, HPA005792). Next, sections were visualized with fluorescent secondary antibodies diluted in PBS-X and then mounted on gelatinized slides.

Light and fluorescence microscopy

Images were acquired using an OLYMPUS BX61VS (VS120-S6-W) scanner (Olympus, Shinjuku-ku, Tokyo, Japan) and ZEISS LSM880 confocal microscope (ZEISS, Oberkochen, Germany). The fluorescent images were acquired in monochrome and color maps were applied to the images post acquisition. Post hoc linear brightness and contrast adjustment were applied uniformly to the image under analysis.

Supplementary Material

Supplementary material is available at PNAS Nexus online.

Competing Interest

The authors declare no competing interests.

Funding

This work was supported by the Brain Science and Brain-like Intelligence Technology - National Science and Technology Major Project (2022ZD0211700) and the National Natural Science Foundation of China grant (32070978) to R.K.N.

Author Contributions

Chao Fang (Investigation, Methodology, Visualization, Writing—original draft, Writing—review & editing), Miao Li (Investigation, Methodology), Lanxiang Li (Investigation, Methodology, Writing—review & editing), Feng Liang (Investigation), Jinfeng Huang (Investigation, Methodology), Li Lu (Resources), Zhonghua Lu (Resources), Hong Wang (Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Validation, Writing—review & editing), and Robert Konrad Naumann (Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Visualization, Writing—original draft, Writing—review & editing)

Data Availability

All data supporting the findings of this study are available within the paper and its Supplementary material.